Stefanie EichertProf. Dr. Jürgen Bauer Department of Dermatology University of Tübingen Liebermeisterstraße 25 D-72076 Tübingen, Germany Tel.: +49-7071-29-84555 Fax: +49-7071-29-5450 E-mail: firstname.lastname@example.org
In recent years, in vivo confocal laser scanning microscopy (CLSM) has become an established method for the non-invasive examination of the skin. In vivo CLSM allows for real-time imaging of micro-anatomic cutaneous structures. It has been used to diagnose ambiguous skin tumors and to measure subclinical tumor spread prior to surgery. By additionally providing high power morphologic information, in vivo CLSM helps to reduce unnecessary biopsies. A multitude of diagnostic features for skin tumors has been published. Here we review published diagnostic in vivo CLSM features, and compare them to our own experience in 100 tumors. In combination with clinical examination and dermatoscopy, in vivo CLSM is a valuable additional tool for non-invasive skin tumor diagnosis.
In recent years in vivo confocal laser scanning microscopy (CLSM) has gained increasing significance as a new method of painless and non-invasive examination of cutaneous tumors at high magnification. It makes real-time viewing of the skin at the level of cellular resolution possible [1, 2]. In comparison to conventional microscopy and dermatoscopy, it has several advantages: confounding information outside the focus is eliminated, the depth to be observed can be selected and it is possible to make a series of section images at one level of depth that approximately corresponds to the thickness of the epidermis. In vivo CLSM can provide additional information on tumors that are unclear in dermatoscopy and thus help elucidate type and dignity of a skin lesion [3–5]. In some cases unnecessary biopsies and surgery can be avoided. The evaluation of the course of unclear tumors can be facilitated and improved by the use of in vivo CLSM in addition to digital dermatoscopy. As the examination does not alter the skin, it can be repeated as desired and dynamic processes can be objectified. Among others, in vivo CLSM is used for preoperative determination of subclinical extension of tumors, for example lentigo maligna , treatment control of topical tumor therapies such as imiquimod 5 % cream [7, 8] or cryotherapy  and superficial postoperative control for recurrence . This article gives insight into possibilities provided by in vivo CLSM and compares already published diagnostic criteria for the different diagnoses with our own experiences.
Principle of in vivo confocal laser scanning microscopy
In confocal laser scanning microscopy, the skin is scanned with a point-like laser beam with a wavelength near the infrared range in vivo. Reflected light traverses a detector with a confocal diaphragm which blocks passage of light from levels over or under the focus level. In the scan only the information of one point in the three-dimensional skin reaches the detector. Thus a high degree of depth discrimination is possible. The skin is scanned in a line-by-line manner and a two-dimensional horizontal image of the confocal level of the epidermis (optical section) is produced [1, 2]. This is repeated for further parallel section levels, so that a three-dimensional volume up to a maximum depth of 0.35 mm can be studied. The individual microstructures of the skin possess differing reflection indices. This provides the image with its natural contrast. The varying reflection of the laser light by each point is measured and consecutively transformed into a digital image with different levels of gray. Melanin reflects strongly (reflection index = 1.7), so that the cytoplasm of melanocytes appears white . Collagen also appears very bright. Keratin reflects less intensely (reflection index = 1.5), so that the cytoplasm of keratinocytes appears somewhat darker. Cell nuclei appear dark. In contrast to light microscopy, the optical sections in CLSM are parallel to the skin surface, i. e. they are oriented horizontally. Depending on type of tissue lateral resolution is 0.4–1.25 μm and the maximum depth of penetration between 200–350 μm [2, 11]. This depth of penetration is usually sufficient to depict the epidermal portion of cutaneous tumors.
In CLSM the histomorphological layers of the epidermis and the arrangement of the rete ridges and the dermal papillae are easily recognized. In normal skin the gray keratinocytes are arranged regularly. The papillae are surrounded by basal cells in a ring-like fashion; depending on skin type and amount of melanin the keratinocytes appear variably bright (Figure 1) [11, 12]. In the center of the papillae, blood vessels are often visible. With the maximum penetration depth of the laser of about 350 μm, one can see into the upper reticular dermis if the epidermis is thin.
Special confocal microscopic diagnostic criteria for the diagnosis of skin tumors with CLSM have been developed. Many alterations visible in light microscopy have a confocal microscopic correlate. In in vivo CLSM in contrast to light microscopy one is dealing with horizontal sections. Further, there are no artifacts caused by fixation, shrinkage or staining. Therefore pathological changes in part appear different than in light microscopy.
Common melanocytic nevi
Common melanocytic nevi have dermatoscopic patterns (globular, reticular, structureless) that can also be recognized well in CLSM . In common melanocytic nevi, the borders of the keratinocytes adjacent to the nevus are easily seen. Monomorphic melanocytes are arranged regularly in the basal layer with accentuation at the tips of the papillae. There is a resemblance to cobblestones. At the dermoepidemal junction, the (dark) papillae are lined by bright monomorphic cells which correspond to keratinocytes with melanin caps as well as melanocytes. These papillae lined in a ring-like fashion are termed “edged papillae”. In reticular melanocytic nevi, the dermatoscopic characteristic network is also seen well in the CLSM image. They possess particularly regularly arranged edged papillae (Figure 2). Globular nevi have dense, homogeneous melanocytic nests, in which the single melanocytes can hardly be distinguished. The papillae are less sharply lined than in the reticular nevus  (Figure 3).
Atypical melanocytic nevi
Atypical nevi can resemble common nevi in CLSM, but they can also reveal changes as seen in melanoma. Their classification is therefore difficult in some cases . Often polymorphic melanocytes, irregularly lined papillae (non-edged papillae) and sometimes dendritic melanocytes that are less highly branched than in melanoma are found . In correlation to dermatoscopy one finds the atypical network and elongated, confluent rete ridges which correspond to the horizontal confluence of nests (bridging) in histology (Figure 4).
Melanoma in situ
The diagnostic CLSM criteria for melanoma in situ and lentigo maligna correspond to those for invasive melanoma . The epidermis appears inhomogeneous and the regular distribution of papillae clearly lined by bright cells is in part disrupted. The melanocytes are polymorphic. Upward-scatter of melanocytes in part with complex branching is seen within higher epidermal layers up to the stratum corneum (Figure 5).
Melanomas are characterized in CLSM especially by polymorphic dendritic melanocytes with complex branching and by non-edged papillae. Striking is an inhomogeneous appearance of the epidermis and an irregular to partially missing honeycomb pattern of the keratinocytes. Pagetoid upward-scattering melanocytes that serve as a diagnostic criterion in light microscopy can also be recognized in CLSM (Figure 6). Invasive melanomas can hardly be distinguished from in situ melanomas, as their invasive character is unreliably depicted due to the limited depth of penetration of the laser .
Basal cell carcinoma
In light microscopy basal cell carcinoma is characterized by typical tumor cell nests with peripheral palisading of the nuclei. They are usually recognized well in CLSM, as long as they are found at the lower edge of the epidermis or in the upper dermis. Blood vessels are dilated and branch in a tree-like fashion . They are striking because they are found directly below the epidermis and usually lie horizontally. In live images of basal cell carcinomas, one can see how large white reflecting cells (leukocytes) move slowly along the vascular endothelium [17, 18] (see online material). Bright collagen fiber bundles surround the tumor parenchyma (Figure 7).
Squamous cell carcinoma and actinic keratosis
The diagnosis of non-melanocytic tumors in CLSM is made more difficult by the fact that the bright contrast of melanin is missing in these tumors. Actinic keratoses and squamous cell carcinomas are characterized by hyperand parakeratoses, large pleomorphic keratinocytes as well as an inhomogeneous epidermis [19, 20]. The borders of the keratinocytes are sometimes poorly defined  (Figure 8).
In some cases a pigmented seborrheic keratosis cannot clinically be definitively distinguished from melanoma. CLSM reveals a specific image resembling gyri and sulci of the brain. Often keratin-filled cyst-like structures are found in correlation with light microscopy. In pigmented seborrheic keratoses, a larger amount of melanin is found in keratinocytes  (Figure 9).
Our own experience with the diagnostic criteria for in vivo CLSM
A multitude of diagnostic criteria for in vivo CLSM [11, 13, 16–24] have been published. We compare these to our own experience and discuss the differences. For this purpose we examined lesions on 100 consecutive patients presenting in our Surgical Dermatology unit for outpatient surgery; included were 56 melanocytic nevi, 11 melanomas, 24 basal cell carcinomas, 8 actinic keratoses and squamous cell carcinomas as well as one seborrheic keratosis (Table 1). Excluded were only tumors that could not be studied on technical grounds and patients for whom organizational demands in the surgical suite did not allow for the about ten minute long examination. All tumors were excised in the framework of routine patient care and sent for histopathology to secure the diagnosis. The confocal microscopic images were analyzed retrospectively with knowledge of the histological diagnosis for the presence of published diagnostic criteria [11, 13, 16–24]. The criteria that achieved the best sensitivities and specificities in larger studies in the literature were examined. For these criteria the published data were compared with the sensitivity and specificity of our cohort. The skin was examined with a VivaScope® 1500 (MAVIG GmbH, Munich, Germany). A diode laser whose wavelength of 830 nm is close to the infrared range was employed. The maximum depth of penetration was 0.35 mm and limited the examination to the epidermis and papillary dermis. To reduce refraction between the skin surface and the objective the optical instrument was coupled to the skin with immersion media. An overview image consisting of 8 × 8 individual images with a size of 500 × 500 μm each at the level of the dermoepidermal junction was made of each tumor. In addition two to four 500 × 500 μm large areas were studied at deeper levels.
Table 1. Studied tumors, male-to-female ratio and median age.
Melanoma in situ/lentigo maligna
Basal cell carcinoma
Squamous cell carcinoma
For common melanocytic nevi the criteria monomorphic melanocytes and edged papillae possessed a specificity of 100 % in our cohort. Well-defined keratinocyte cell borders were present in 72.7 % of the nevi, monomorphic melanocytes in 70.9 %. Good sensitivity was achieved with the combination of both criteria (Table 2). The melanoma criteria polymorphic melanocytes, inhomogeneous epidermis as well as non-edged papillae were found in all melanomas studied. Complexly branching dendritic cells were a highly specific parameter with 97.2 %. Polymorphic melanocytes had a specificity of 83.3 % (Table 2). In the basal cell carcinomas increased, enlarged blood vessels were observed regularly (sensitivity 95.8 %). Bright collagen fiber bundles in contrast were seen less often (83.3 %) and were as the other criteria, too, more unspecific (62.5 %). Long monomorphic cells and the parallel polarization were detected only irregularly. Peripheral palisading of nuclei in dark cell nests was present in 79.2 % of basal cell carcinomas and highly specific (100 %) (Table 3). The published criteria for actinic keratoses and squamous cell carcinomas were present in every tumor in our cohort, i. e. the sensitivity was 100 % in each case. Hyperkeratoses also had a specificity of 100 %. Atypical keratinocytes and irregular keratinocyte cell boundaries each had a specificity of 91.7 % (Table 3).
Table 2. Diagnostic in vivo CLSM features of melanocytic skin tumors.
Monomorphic, bright, round to oval cells with well-defined cell borders
In vivo confocal laser scanning microscopy is a new, non-invasive method that makes an “optical biopsy” of the skin possible. Numerous publications describe diagnostic criteria for melanocytic and non-melanocytic tumors that are suitable for the evaluation of tumors, where clinical features and dermatoscopy do not allow for definitive diagnosis. The 100 tumors studied for this paper demonstrate the utility of this method in daily clinical routine, even if time-consuming. They also showed that published criteria allow for good diagnostic classification of the tumors. Nevertheless, the sensitivities and specificities in our study must be interpreted cautiously due to the small case number and be confirmed by larger studies.
Common melanocytic nevi can easily be recognized by CLSM. It can be difficult to distinguish atypical melanocytic nevi from melanoma, as similar criteria in varying intensity apply to both. In both, for example, often no edged papillae or monomorphic melanocytes are detected. As more than half of our collective of 56 nevi were atypical, the sensitivity of the diagnostic criteria for melanocytic nevi was comparatively low (Table 2). In contrast, the sensitivity of the nevi criteria in differentiation from melanomas was high (specificity 81.8 %–100 %). In the literature the criterion of monomorphic melanocytes was reported as highly sensitive with 98.9 % and as highly specific with 98.1 %. In this study, however, only few images per tumor were made, so that possibly a certain selection of characteristic alterations existed. Homogeneous nests were seen in only 47.3 % of our nevi, as they do not occur in reticular nevi and only occur as irregular nests in dysplastic nevi. For the diagnosis of melanocytic nevi in our experience the criteria monomophic melanocytes, well-defined keratinocyte cell borders and edged papillae are suitable.
The 11 melanomas in our cohort could be reliably identified with the published criteria (Table 2). The criteria inhomogeneous epidermis and irregular honeycomb pattern refer to disturbed epidermal architecture and are used largely synonymously in the publications. We found the criterion inhomogeneous epidermis to be highly sensitive (100 %) and highly specific (94.4 %). Polymorphic melanocytes and non-edged papillae were found almost universally in published studies and in our cohort (sensitivity 94.6 %–100 %). The lower specificity of both criteria is explained by the fact that they are also frequently found in the atypical nevus. Complexly branched dendritic cells with 99.4 % in the literature and 97.2 % in our study are highly specific for melanoma. Nevertheless, they were detected in only 6 of 11 melanomas (54.5 %). Upward scatter of melanocytes was also not very sensitive (54.5 %) but highly specific (89.3 %) for melanoma. In a study on 37 melanomas and 88 melanocytic nevi with the combined use of CLSM and dermatoscopy every melanoma was identified.  In lightly pigmented or ame-lanotic melanomas the specificity of CLSM is significantly higher (84 %) than dermatoscopy (39 %) [10, 25]. In view of published and our own data we propose four diagnostic criteria: inhomogeneous epidermis, polymorphic melanocytes, complexly branched den-dritic cells and non-edged papillae. By combining dermatoscopy with CLSM a higher sensitivity can be achieved than with dermatoscopy alone .
Basal cell carcinomas have characteristic features in CLSM. An increase of atypical ectatic blood vessels is a highly sensitive criterion for basal cell carcinomas [16, 18, 21]. In our cohort of 24 tumors sensitivity also was 95.8 %. The specificity of the criteria was calculated in comparison to squamous cell carcinoma. In our study none of the published criteria was particularly specific for basal cell carcinoma, as, for example, enlarged blood vessels and elongated cells with parallel distribution also were present in squamous cell carcinomas. We examined the basal cell carcinomas for the characteristic criterion in light microscopy peripheral palisading of elongated nuclei in dark cell nests and found this criterion in 79.2 % of basal cell carcinomas. This feature was not found in any other tumor (specificity 100 %) and was observed in 2 of 4 sclerosing basal cell carcinomas in our study, while here less nests but more strands partially with peripheral palisading of elongated nuclei were present. In a smaller study with 14 basal cell carcinomas peripheral palisading was observed in only five tumors . Bright collagen fiber bundles were seen in 83.3 % of basal cell carcinomas, but had in comparison to squamous cell carcinomas only a specificity of 62.5 %.
Suitable for the diagnosis of basal cell carcinoma in our experience are the criteria of increased, enlarged blood vessels and palisading of elongated nuclei in dark cell nests. For the examination of superficial subclinical spread  and equally superficial postsurgical control for recurrence as well as control of topical therapy with immunomodulators [7, 8], CLSM can be an aid. Due to the limited depth of penetration, CLSM cannot determine if topical therapy was effective deeper than 350 μm or if a recurrence is present there. Consequently, in therapy and recurrence control CLSM can lead to false-negative results.
The same diagnostic criteria apply to actinic keratoses and squamous cell carcinoma. Both possess identical morphological changes in their superficial portions that can be evaluated with CLSM [26, 27]. The four criteria employed were present in our cohort in all 3 actinic keratoses and 5 squamous cell carcinomas (sensitivity 100 %). Specificity in our study in differentiation from basal cell carcinomas was between 79.2 % and 100 % (Table 3). In the literature the individual criteria are described as considerably less specific which might be explained by the fact that there they were compared to normal skin [19, 20]. The criterion hyperkeratosis had the highest specificity of 100 % in our study, while in another smaller study on seven patients it was only 50 %. Here only six healthy patients were examined, of which three had actinically damaged skin and thus also hyperkeratoses. In a study on 44 actinic keratoses all four criteria were combined and thus 97.7 % of the actinic keratoses were identified . Here CLSM is described as a diagnostic possibility to distinguish actinic keratoses from normal skin and to detect subclinical alterations. In a recent study 10 actinic keratoses were examined with CLSM before and 3 and 12 months each after shave biopsy. Initially and after 12 months 10 of 10 actinic keratoses were correctly identified, after 3 months 8 of 10 . Here, too, the authors consider CLSM as a useful method for follow-up of clinically and dermatoscopically unclear findings after surgical interventions.
Suitable diagnostic criteria for actinic keratoses and squamous cell carcinomas in our experience are hyperkeratoses, irregular honeycomb pattern, atypical keratinocytes and poorly defined keratinocyte cell borders. Hyperkeratoses also hinder confocal microscopic diagnostics as strong cornification in part prevent applying the microscope flat on the tumor, so that air bubbles develop in the image. Further, the penetration depth of the laser does not reach the altered vital layer, mainly the stratum spinosum . This usually also makes it impossible to definitively rule out invasion and thus differentiate an actinic keratosis from a squamous cell carcinoma securely. In published studies and in the present study highly hyperkeratotic tumors were therefore excluded.
A limiting factor for the examination of skin tumors with in vivo CLSM is the long duration of the examination of about 7 to 10 minutes. Further, tumors on uneven body sites such as the nose or helix of the ear can hardly be examined, as the objective cannot be applied stably and flatly at these sites. Equally, an image is impossible in case of extensively ulcerated or highly hyperkeratotic tumors. Another problem is presented by tumors in locations with a thick epidermis such as palms of hands and soles of feet. The depth of penetration of the laser is insufficient at those sites to capture relevant alterations . Further, the high costs of in vivo CLSM devices – VivaScope® 1500 (MAVIG GmbH, Munich, Germany) costs about 80,000 € at present – still restricts wide distribution of this method in clinical practice.
As a supplement to dermatoscopy besides in vivo CLSM in recent years several other high-resolution techniques have been developed for dermatologic in vivo diagnostics. For example, high-resolution ultrasound is employed for the diagnosis of melanoma metastases  and for preoperative measurement of tumor thickness . However, it has not become established for the early diagnosis of skin cancer due to the superficial location and low depth of penetration of the tumors. Spectrophotometric analyses (SIAscopy: spectrophotometric intracu-taneous analysis) are also used to improve diagnostics of cutaneous tumors. In two recently published studies on melanoma  and pigmented basal cell carcinomas  it did not prove itself in comparison to dermatoscopy. Optical coherence tomography (OCT) allows for vertical sections up to a depth of several millimeters. Its use in the diagnostics of skin tumors is the subject of current, intensive research . The morphology of individual tumor cells cannot be determined with this method.
In conclusion, in vivo CLSM is a promising and elegant method for supplemental, high-resolution diagnostics and follow-up of skin tumors, particularly when used in conjunction with dermatoscopy. Especially lightly pigmented melanocytic tumors that possess few clinical and dermatoscopic diagnostic structures are depicted well in in vivo CLSM.